Marcinko Charlotte L J, Samanta Sourav, Basu Oindrila, Harfoot Andy, Hornby Duncan D, Hutton Craig W, Pal Sudipa, Watmough Gary R
School of Engineering, University of Southampton, Southampton, SO17 1BJ, UK.
School of Oceanographic Studies, Jadavpur University, 188, Raja S.C. Mallik Road, Jadavpur, Kolkata, 700032, India.
J Environ Manage. 2022 Jul 1;313:114950. doi: 10.1016/j.jenvman.2022.114950. Epub 2022 Apr 1.
There is increasing interest in leveraging Earth Observation (EO) and geospatial data to predict and map aspects of socioeconomic conditions to support survey and census activities. This is particularly relevant for the frequent monitoring required to assess progress towards the UNs' Sustainable Development Goals (SDGs). The Sundarban Biosphere Reserve (SBR) is a region of international ecological importance, containing the Indian portion of the world's largest mangrove forest. The region is densely populated and home to over 4.4 million people, many living in chronic poverty with a strong dependence on nature-based rural livelihoods. Such livelihoods are vulnerable to frequent natural hazards including cyclone landfall and storm surges. In this study we examine associations between environmental variables derived from EO and geospatial data with a village level multidimensional poverty metric using random forest machine learning, to provide evidence in support of policy formulation in the field of poverty reduction. We find that environmental variables can predict up to 78% of the relative distribution of the poorest villages within the SBR. Exposure to cyclone hazard was the most important variable for prediction of poverty. The poorest villages were associated with relatively small areas of rural settlement (<∼30%), large areas of agricultural land (>∼50%) and moderate to high cyclone hazard. The poorest villages were also associated with less productive agricultural land than the wealthiest. Analysis suggests villages with access to more diverse livelihood options, and a smaller dependence on agriculture may be more resilient to cyclone hazard. This study contributes to the understanding of poverty-environment dynamics within Low-and middle-income countries and the associations found can inform policy linked to socio-environmental scenarios within the SBR and potentially support monitoring of work towards SDG1 (No Poverty) across the region.
利用地球观测（EO）和地理空间数据来预测和绘制社会经济状况的各个方面，以支持调查和普查活动，这一兴趣正在与日俱增。这对于评估联合国可持续发展目标（SDGs）进展所需的频繁监测尤为重要。孙德尔本斯生物圈保护区（SBR）是具有国际生态重要性的地区，拥有世界上最大红树林的印度部分。该地区人口密集，有超过440万人居住，许多人生活在长期贫困中，严重依赖基于自然的农村生计。这种生计容易受到包括气旋登陆和风暴潮在内的频繁自然灾害的影响。在本研究中，我们使用随机森林机器学习方法，研究了从EO和地理空间数据得出的环境变量与村庄层面多维贫困指标之间的关联，以提供支持减贫领域政策制定的证据。我们发现，环境变量能够预测SBR内最贫困村庄相对分布的78%。气旋灾害暴露是预测贫困的最重要变量。最贫困的村庄与相对较小面积的农村定居点（<约30%）、大面积的农业用地（>约50%）以及中度到高度的气旋灾害相关。最贫困的村庄的农业用地生产力也低于最富裕的村庄。分析表明，拥有更多样化生计选择且对农业依赖较小的村庄可能对气旋灾害更具抵御能力。本研究有助于理解低收入和中等收入国家内的贫困与环境动态关系，所发现的关联可为SBR内与社会环境情景相关的政策提供参考，并有可能支持对该地区实现可持续发展目标1（消除贫困）工作的监测。
